alt Hacker NewsClaude in particular has nothing to do with it. I see many people are discovering the well-known fundamental biases and phenomena in LLMs again and again. There are many of those. The best intuition is treating the context as "kind of but not quite" an associative memory, instead of a sequence or a text file with tokens. This is vaguely similar to what humans are good and bad at, and makes it obvious what is easy and hard for the model, especially when the context is already complex.
Easy: pulling the info by association with your request, especially if the only thing it needs is repeating. Doing this becomes increasingly harder if the necessary info is scattered all over the context and the pieces are separated by a lot of tokens in between, so you'd better group your stuff - similar should stick to similar.
Unreliable: Exact ordering of items. Exact attribution (the issue in OP). Precise enumeration of ALL same-type entities that exist in the context. Negations. Recalling stuff in the middle of long pieces without clear demarcation and the context itself (lost-in-the-middle).
Hard: distinguishing between the info in the context and its own knowledge. Breaking the fixation on facts in the context (pink elephant effect).
Very hard: untangling deep dependency graphs. Non-reasoning models will likely not be able to reduce the graph in time and will stay oblivious to the outcome. Reasoning models can disentangle deeper dependencies, but only in case the reasoning chain is not overwhelmed. Deep nesting is also pretty hard for this reason, however most models are optimized for code nowadays and this somewhat masks the issue.
Replies
Author here, yeah I think I changed my mind after reading all the comments here that this is related to the harness. The interesting interaction with the harness is that Claude effectively authorizes tool use in a non intuitive way.
So "please deploy" or "tear it down" makes it overconfident in using destructive tools, as if the user had very explicitly authorized something, and this makes it a worse bug when using Claude code over a chat interface without tool calling where it's usually just amusing to see
So easy it should disqualify you if you fail this: Knowing your own name.
You can really see this in the recent video generation where they try to incorporate text-to-speech into the video. All the tokens flying around, all the video data, all the context of all human knowledge ever put into bytes ingested into it, and the systems still completely routinely (from what I can tell) fails to put the speech in the right mouth even with explicit instruction and all the "common sense" making it obvious who is saying what.
There was some chatter yesterday on HN about the very strange capability frontier these models have and this is one of the biggest ones I can think of... a model that de novo, from scratch is generating megabyte upon megabyte of really quite good video information that at the same time is often unclear on the idea that a knock-knock joke does not start with the exact same person saying "Knock knock? Who's there?" in one utterance.